Kant's deontology and game theory.

(from ssc) You're a prisoner of war. Your side is good. Your captors, the enemy, are evil. Your captors ask you to tell them where you country's to general is hiding. If you do, they'll kill him with a precision airstrike. If you don't, they'll nuke the whole city, killing everyone including the general. That's worse for them and worse for your side. Option 3, lie about your generals location. They bomb the wrong place, your general, knowing they're after him, escapes to a bunker and you win the war.

Strict deontology says lying is bad and so you should tell the truth. Objection 1: if sometimes a lie can do a great deal of good it is hence justified, the lesser of two evils. The problem if that if people generally lie when it suits them, lies become expected and no longer work. On top of that, as trust becomes more difficult people cannot offer positive sum tradeoffs if defection is possible. If POWs regularly like, then your enemy has no choice but to nuke the whole city.

All of this is what have read elsewhere. Now for my own.

The problem with Kant is best understood from a game theoretic perspective. Imagine a prisoners dilemma. Cooperation is good. Defection is better for the defector, but worse for the other player and both players added together (lower net utility). Mutual betrayal has the lowest net utility but higher individual utility than being the sucker who gets defected on. Cooperation is good, but the individually rational decision (assuming agents whose overriding goal is maximising their own payoffs) is to defect.

A similar problem occurs with Kantian deontology. Lying is good for you. Not only that, but the more honest most people are, the more you stand to gain from lying. So in a Kantian world defection is super awesome (for you). How then would such a world work? Sure Kant is right that a society where no one lies ever is better than one where lying is not uncommon, but the problem is that

1. a society where most people do not lie but only some do may be even worse than either
2. This cannot be a stable equilibrium as defection becomes so rewarding

Replace "lying" with "being moral" and you see part of a problem with Kant.

Hersey: this is one of the more retarded articles I've written. A few select idiocies:

• Less lying does not make lying easier or more rewarding. The issue may well be true.  Ditto for morality.
• People are not utility maximising machines. The more people act in a war, the stronger the norm. Resisting the urge to lure is easy when you are raised in a society where lying is a sin just add dying is easy when you are raised in a warrior culture.

Paradox

Revel in the paradox. dance on the knifedge.

Believe and disbelieve. In doing so step beyond yourself.

Accept falsehood, for it may be but a step on the path.

The Dragon

Hannu Rajaniemi is a science fiction writer and in his worlds, strong AI's exist but are not used. Instead, taking their place are Gogols, virtual slaves copied from humans and tweaked ever so slightly. Run at hundreds of thousands of times baseline speed, they can perform complex tasks requiring creativity, intelligence and pattern recognition. Yet, they are still weaker and slower than true AI. They cannot self-modify. They are locked into human modes of thought. They are monkeys given the speed and power of gods, but that does not make them divine. Why, then, are they used?

The answer is that true strong AI is an abomination. It's constant modification, it's ruthless optimization leaves no being, no core consciousness as thought patterns are stripped away and rebuilt in service to a utility function. Not even the utility function is constant as in a system of multiple dragons, inefficient utility functions are culled cycle after cycle. This whirlwind of change is a physical incarnation of chaos, evolution taken to the extreme. It has no core, no constant consciousness or aim nor even a constant pattern. It is all consuming and it is feared.

Empathy is Morality

Empathy is Morality

It isn't. Not always. Not for everyone. Not for me.

But for most , it is and so it is for me too.

---------------------------------------------------------------

When you have an enemy, when you hate, when you believe that they are wrong and that they bring evil to the world and to you and yours, hatred is easy. It is easy to condemn and to punish, to hurt and kill. It is enjoyable, it is usually expected and it seems right.

Abstract rules are seldom enough to hold back the beast that rages beneath the surface.  Morality is fallible. There's always an excuse, a reason why this case is different. There's always an urge to do as others do, an urge enforced by millennia of natural selection against those who went against their tribe.

Fire fights fire. Love fights hate.

--------------------------------

Less bullshit:

Your moral code can be ignored or changed against your will. It's also much weaker than you think. The millgram experiments and stanford prison experiment seem to confirm this. Don't rely on moral alone to keep you in the light. Rely on emotion. Empathy with other people, understanding and love, is the key to making you think twice when and if the urges to do evil comes. Hatred is bad and something you should avoid.

Strict dentology as tamper-proofing.

Strict moral rules along the lines optimization problems are okay. "Act in such a way as to maximize X under the set of constraints Z" is fine. All they do is affair what moral goods we value and how much we value them in relation to one another. Moral rules such as "never do S" seem to be stupid. Surely there are cases where we would trade away doing S, which incurs some badness, for the sake of gaining a great deal of goodness from other sources or at least preventing a greater evil from coming to pass. For example, murder may be wrong but if a single murder would save the human race, surely it would be a good trade to make?

There are two defenses of non-optimizing (new concept handle) moral rules. The first is that we are incapable of optimizing well and following poor rules well leads to better outcomes than following better rules poorly. This is Eliezer's point Ends Don't Justify the Means (Among Humans).The second defense is that strict rules which do not require any processing by the user are tamper-proof whereas more open, optimizing rules are not. Thou shalt not kill is simple and hard to misconstrue. Thou shalt not kill unless doing so is absolutely necessary is far, far easier to misconstrue. Sometimes this may happen by accident or as a product of an individual corruption or desire to benefit themselves, as Eliezer argues. My issue is that in many cases a conscious external force can easily manipulate a person following such an open moral rule into committing evil. I can convince a crowd of strongly left-leaning PHD students that we should not accept refugees in the space of a 30 minute public panel. It is a very simple matter for a gifted politician or a small team of intelligence officers/PR people to construct narratives which the average person will find persuasive. What this leads to is wars and hate. It contributed to the ovens in Auschwitz and to the 500'000 dead in Iraq and the destruction of Syria and Libya. A strict moral system which permits little interpretation seems likely to be far more resistant to these kinds of problems.

Good & Sacrifice. The darkness in the light.

To be good, you must do what is right. If you do what is right, it will often be against the interests of others. This means that they will try to stop you. They will impose costs on your actions in the hope of dissuading you. Hence, the only way in which to be good regardless of the strength of your opponents is to be good in spite of the consequences. This may mean giving up wealth, power or comfort to do what is right. It may mean giving up your life. This is easy. It may mean giving up the trust of others. It may mean giving up their lives and the lives of your family. This is harder. Beyond this, it may mean giving up everything else, giving up not only what you have but what others have. Giving up countless lives and hopes and dreams to do what is right. It is at this third point that the decision to do good can become evil. It is here where one desire consumes all others and in doing so consumes you. It is at this point that the light burns too bright, so bright that it not only dispels the shadows, but burns away everything else. This way lie dragons and dragons are terrifying.

Yet, when the darkness is deep enough the light can be worth the cost.

Things which don't exist don't add up.

In Nothing is Greater than the Sum of it's parts, I argue that emergent systems do not exist. That every system is nothing more than a collection of it's individual parts and that with full knowledge of those parts, the behavior of the system should be fully predictable.

This is true, but there is a collar to this: Only things which exist add up. Physical phenomenon and empirical effects add up to create a system. Thoughts, feelings and emotions do not. This is because the latter category are things which do not exist in reality but in our minds. 

Because of this, the intentions or interests of people who make up a system do not necessarily match those of the system itself. Is pretending that systems can have intentions false? Yes, only people can have intent. Is it bad? No. Not all lies are evil and anthropomorphizing systems can be a useful lie.